A new method is described of deriving a distribution, P(x), in a set of variables, x, which is consistent with a set of data, A (x), and which is the least biased modification (maximizes entropy) of a constraining distribution, r(x). The method is demonstrated using nuclear spin dipolar couplings obtained for n-alkanes dissolved in, a Liquid crystalline solvent. The constraining distributions are obtained from computer simulations of the molecular dynamics.